Also known as Foreground Detection is a widely known technique to extract image's foreground

Wikipedia: Background Subtraction

Implementation Details

This project implements the 1-D Gaussian background subtraction algorithm. Each individual pixel is modeled using a Gaussian distribution.

Notes

  • Binary thresholding is applied by setting foreground to 255 (white) and background to 0 (black).
  • Initial threshold (k) and temporal window (𝜌) values are set based on trial and error.

Source Code:

GitHub

Instructions to Run

  1. Install and configure OpenCV.
  2. Download image sequence from here: Office Dataset
  3. Extract images into Images directory. (Image file names format: inxxxxxx.jpg)
  4. First 619 images are used for generating binary threshold version of image in000620.jpg. Adjust variables as need be.
  5. Compile and run.